Flexible Transit Network Design With Variations Of Bus Travel Demand

The transit demand varies with time, and both the demand amount and the distribution of transit demand will be different in different periods. The supply of transit network capacity meeting the change of transit demand, and achieving the balance between demand and supply in different periods, is essential to enhance the transit system service level, to reduce the waste of public transportation resources, and to improve the interests of society as a whole. At present, the design of the bus network is based on single bus demand OD table, and the lack of the consideration of bus demand change results in the difficulty to respond to the bus demand change effectively for bus network.In this paper, analyzing the law of bus demand change of long-term and short-term, finding the obvious regularity of short-term changes of bus demand, and the short-term change can be induced as fluctuations change process one week as a cycle, the weekly changes in demand can be reflected by the OD table of seven days. The periodic fluctuation of the change of bus demand lay the basis for the design of the bus network scientifically. In order to deal with the change of bus demand, the paper proposes a method of building flexible bus network to deal with the change of bus demand. After giving the concept of flexible bus network, analyzes the structure of the flexible bus network, dividing the bus network flexibility into three levels including the buffer flexibility, the adaptation flexibility, and the innovation flexibility, reflecting the ability of understanding demand, meeting demand, guiding demand respectively. Among which the buffer flexibility is the basis for building the flexible bus network.The emphasis of the paper is the method of building the buffer flexibility of bus network. Analyzing the interests of the three parties including the passengers, the bus companies and the society, regarding the weighted sum of the above three parties’ interests as the designing goal of bus network, making the variable bus OD demand table as the design foundation, proposing the objective function of bust network design, building a multi-objective optimization model for designing bus network scientifically. The model includes six optimization objects, such as passengers’ total waiting time, overload operation time, unused capacity time, bus fleet size, passengers’ total increased time than ideal network and the amount of unsatisfied passenger, using these indicators to consider the interests of all three parties, calculating the objective function value of bus network under different demand patterns, using the objective function value as indicator reflecting the bus network’s fitness to different demand patterns, and the fitness are quantified by money amount. The optimal bus network are founded by solving the model, enhancing the buffer flexibility of the bus network, and improving the ability of the bus network to deal with the changes of bus demand.The solving of the bus network design model is a multi-objective optimization problem, and a decomposition algorithm is proposed to reduce the complexity of the model solving. At first, the algorithm selects the candidate bus lines to reduce the search space, and at the second step, a genetic algorithm is used as the search algorithm. At the first step, the algorithm reduces solution space at the bus line level, at the second step, the genetic algorithm is applied to find the optimal combination of the bus lines at network level, and optimal chromosome selection and simulated annealing tension algorithm are introduced into the genetic algorithm as two improvement strategies to ensure the ideal direction of evolution of the population. The results of the case show that the bus network can better adapt to the changes of the bus demand and the bus network is of more buffer flexibility when considering the change of bus demand during the network design process. After the screening of the candidate bus lines, the search space of the bus lines can be reduced obviously, and the convergence speed of genetic algorithm can be accelerated largely.